1. Introduction
Food waste is produced in large quantities in industrial production, agricultural processing, domestic consumption, and retail. It is one of the world’s most significant problems in both wealthy and developing nations. Every year, over one billion tons of food is lost or thrown away, accounting for one third of all the food produced for human use, resulting in significant social, economic, and environmental problems [
1]. In terms of the total municipal solid waste yearly (MSW) produced in ASEAN, Thailand has the second highest amount of MSW (26.77 million tonnes/year), while the first is from Indonesia, with 64 million tonnes/year, and in third place is Vietnam, with 22 million tonnes/year [
2]. The food industry generates a large amount of by-products, which causes substantial economic, managerial, and environmental issues, in addition to a significant loss of precious resources. The fruit processing sector falls under the category of food processing. International law addresses the disposal of fruit industry by-products (peels, seeds, and oilseed meals) as a difficulty. Finding new prospective application areas for these wastes to further utilize them for the creation of high-value goods has therefore received increased attention. Currently, the fruit processing industry discards its by-products (peel and seed) because they have no economic use. Because of the scarcity of landfills and the high cost of transportation, the disposal of these wastes incurs extra costs for the fruit processing industry. Due to their high chemical oxygen demand (COD) and biological oxygen demand (BOD), as well as their rapid rate of disintegration, which provides an ideal environment for the reproduction of insects, they should be disposed of carefully.
Both manufacturers and retailers are likely to throw away unconsumed processed and fresh foods. This amount of waste from agricultural production, including the manufacturing and retail industries, is classified by a proportion of 50%, which is made up of raw and cooked peel and seed. According to the United Nations Food and Agriculture Organization, the breakdown of logistical and infrastructural systems causes some countries to waste 55 percent of their agricultural output [
3]. Whether pickled, preserved, or dried, these processed products cause problems with waste disposal and affect pollution in terms of smells and community cleanliness, which cannot be dealt with efficiently [
4,
5]. Based on previous research [
6,
7,
8,
9], the development of agri-waste management platforms could still be novel, leaving a research gap in the development of digital platforms as innovative tools for agricultural waste management in Thailand.
Moreover, the circular economy is used to solve food waste, by converting this waste into high-value-added products [
10]. Hence, developing agri-waste digital platforms to use as an intermediary for the purchase of scraps from fruit/vegetable processing can help to reduce the agricultural waste issues in northeastern Thailand. This northeast region contains 46% of the agricultural holdings in Thailand and 47% of its farm area, with an average holding size of 3.2 hectares. Wastes, in this case, may come from fruits such as mangoes, oranges, pomelos, coconuts, limes, corns, and sugarcanes, and vegetables such as tomatoes, lettuces, kales, and cabbages. Farmers or sellers play a significant role in the platform by bringing the waste to be sold using this platform. Still, most of the previous studies on digital platforms as an innovative tool for managing food waste have focused on technology and innovation to develop digital platforms, leaving a second research gap in exploring behavioral intentions in a deeper dimension with regard to digital platforms [
11,
12,
13]. Previous research related to circular-economy-based digital technology has included process-based information systems (PBIS), which help with digital tracking, modeling, and sensing for recycling, reuse, and remanufacturing [
14,
15].
Therefore, developing a digital platform can help solve inefficient agri-waste management by collecting and managing waste data and providing possible business partners that could use agricultural waste as a high-value-added product in the future. The swimlane diagram of this digital platform is demonstrated in
Figure 1. This digital platform helps to gather the network of vegetable and fruit retailers in fresh markets and processed food manufacturers that have food wastes such as peels, seeds, and flesh. It allows businesses that need materials transformed from these wastes (such as paper, bioplastic, vegan leather, plywood, and straws, among others) to order their value-added materials/products from this platform. The biomaterial pilot plant acts in the middle and is responsible for transforming these agricultural wastes into value-added materials/products. This platform also helps to manage the monetary transaction from the beginning, including collecting money from the business customers, distributing money to waste sellers, managing logistics, and the final stage, which is to send the materials/products to the business customers.
However, the idea of a circular economy is still relatively novel in Thailand, particularly among food merchants and farmers. It is necessary to have a profound understanding of the individuals that would use this technology. Based on previous research, digital platform developers can better comprehend user behaviors when categorizing these users into specific segments and analyzing the factors affecting their use intention [
9,
16]. However, only a small number of studies have focused on the factors influencing these behavioral intentions and the segmentation of digital platform users for a circular economy. In addition, this study contributes to the current body of the literature by analyzing the moderating effect of demographic segments on a number of factors that influence behavioral intentions. On the basis of their gender, age, and annual income, the users were divided into two distinct demographic subgroups. The research framework was based on the Unified Theory of the Adoption and Utilization of Technology (UTAUT2), which posits the determinants of technology adoption, including performance expectancy, effort expectancy, social influence, facilitating condition, hedonic motivation, price value, and habit [
17]. The UTAUT2 was justifiable in this context because evidence revealed that the UTAUT2 has been employed in research on food and agricultural technology to predict users’ behavioral intentions [
18,
19,
20]. An additional academic contribution is that the UTAUT2 can be expanded with more variables, such as trust and privacy [
21], to enrich the research framework and provide novelty for technology adoption concerning circular-economy-based digital technology.
With this novel platform, this study contributes to bridging the gap of factors that influence users’ behavioral intentions toward a digital waste management platform. Explicitly, this study aimed to deeply understand the digital adoption behaviors of users who are vegetable and fruit retailers and processed food retailers. First, we classified the users using the multivariate demographic segmentation approach. Then, we employed the multigroup structural equation modeling approach to explore the factors influencing their behavioral intentions towards adopting digital technology based on the UTAUT2 model.
4. Results of the Study
4.1. Multivariate Demographic Segmentation
According to the cluster analysis results from
Table 1, we divided the users into Segment 1 (
n = 122) and Segment 2 (
n = 204). The chi-square test demonstrated that gender was an insignificant variable for the demographic segmentation (
p-value = 0.163 > 0.01). Most respondents in Segment 1 were people from Generation X and baby boomers. Furthermore, most people in this group had low incomes and were fruit and vegetable retailers, but the higher-income earners were manufacturing officials. However, in Segment 2, Gen Z and Gen Y members made up the bulk of the participants, who clearly earned less than BHT 15,000. This result concluded that the users were classified into two different groups: (1) older people with various income ranges, and (2) young people with low incomes, where gender was ineffective in classifying these segments. As for Segment 1, the respondents’ monthly incomes were less than BHT 15,000.
Next, the perception scores of Segment 1 (older people with various income ranges) and Segment 2 (young people with low incomes) were compared.
Table 2 shows that, compared to Segment 1, the users in Segment 2 had higher mean scores across the board with regard to performance expectancy, effort expectancy, social influence, habit, and behavioral intentions. In contrast, the Segment 1 respondents averaged a greater pricing value and a higher level of privacy than the Segment 2 respondents. However, regarding facilitating conditions and hedonic motivation, comparing the mean scores of Segments 1 and 2 was inconclusive.
After applying
t-tests across the two groups, there was a statistically significant difference between the groups’ mean scores, given
p-values of <0.05, <0.01, and <0.001.
Table 2 demonstrates that Segment 2 tended to have a greater perception that the platform may increase productivity (PE3) than Segment 1. Segment 2 also expected the platform to be easier to learn than Segment 1 (EE1). Moreover, the perceptions of social influence regarding referrals (SI1), influencers (SI2), and respected people (SI3) in Segment 2 seemed to be better off than those in Segment 1. The perception of habits, such as addiction (HB2) and routine (HB3), in Segment 2 tended to be higher than those in Segment 1. Segment 2 also expected to have the necessary knowledge to utilize the platform more than Segment 1 (FC2). In contrast, the
t-test results show that Segment 1 expected to have more support resources to help them utilize the platform (FC1) and required more privacy (PR4) than Segment 2.
There are two primary steps in conducting a statistical test with SEM: the measurement model (CFA) and the structural model [
123].
4.2. Confirmatory Factor Analysis (CFA)
A CFA was utilized in order to conduct tests on the measurement model. Within the scope of this investigation, the model was evaluated with regard to its internal consistency, convergent validity, and discriminant validity. The CFA was carried out by coupling all the constructs to their respective covariances. Before testing could begin, every construct had to have its own set of manifest variables. The goodness of fit (GOF) of the entire connection might be developed based on the covariances among the errors that occurred within the same constructs. This analysis can be elaborated as follows.
4.2.1. The Goodness of Fit (GOF)
In
Table 3, we see the GOF measurements and thresholds. All the metrics met or exceeded the established standards; hence, the outcomes were acceptable. The values for the comparative fit index (CFI; 0.913), incremental fit index (IFI; 0.0914), TLI Tucker–Lewis index (TLI; 0.900), and root mean square error of approximation (RMSEA; 0.071) all exceeded their predetermined thresholds for goodness of fit. In terms of the GOF condition, the stated thresholds were a CMIN/df less than 3.00, CFI larger than 0.900, IFI greater than 0.900, TLI greater than 0.900, and RMSEA less than 0.100.
4.2.2. Convergent Validity
A comparison of the model’s outcomes with the predetermined thresholds for the fit index was used to analyze the convergent validity. Cronbach’s alphas, average variance extracted (AVE), and composite reliability (CR) were utilized to analyze the measures’ degrees of consistency against the recommended thresholds for an AVE and CR of 0.05 and 0.70, respectively. Accordingly, the thresholds for the convergent validity measurements and the derived indicators are listed as follows.
Table 4 shows that, when the calculated measures were compared to the thresholds, the values for the performance expectancy, effort expectancy, social influence, facilitating condition, hedonic motivation, price value, habit, trust, privacy, and behavioral intention all passed the convergent validity criteria. In terms of the constructs, each and every indicator reached a level of statistical significance of <0.001, and all the AVEs were greater than the thresholds (AVE > 0.50). On the other hand, the Cronbach’s alphas and CR values were all higher than 0.7, which indicated that all the indicators included in this measurement model met the criteria for the convergent validity.
4.2.3. Discriminant Validity
The discriminant validity of a test or measurement is defined as the degree to which it varies from another test or measurement with the same underlying idea. To evaluate this component, the Fornell and Larcker criteria were utilized to compare the square root AVEs (located on the diagonal) and the correlations of the various matrices.
Table 5 shows that the square root of each bolded AVE was bigger than the off-diagonal correlation coefficients. This implies that all of the variable compositions may theoretically measure the unique constructs, and the outcome of this was satisfactory. Furthermore, each AVE’s square root was greater than the off-diagonal correlation coefficients.
In addition, Henseler et al. (2015) used the heterotrait–monotrait (HTMT) ratio technique to examine discriminant validity [
124]. This was performed because the Fornell and Larcker (1981) criterion was criticized for its lack of reliability in addressing the distinctiveness between latent variables [
124]. Specifically, this problem was addressed by Henseler et al. (2015). HTMT values greater than 0.90 suggest the presence of discriminant validity between the associated latent variables. According to
Table 5, each value on the HTMT is lower than 0.90, indicating that it satisfies the criteria for the discriminant validity.
4.3. Structural Model
We come to demonstrate the SEM analysis after achieving the prerequisite for the reliability and validity. According to
Table 6, as Hu and Bentler (2009) suggested, the goodness of fit criteria mostly supported this structural model [
125]. According to
Table 7, the structural model’s test results supported H3, H4, and H7 to H9 at a significant level of 0.05 or less. This showed that the correlations between the constructs were statistically significant. The analysis was based on the following concepts: performance expectancy (PE), effort expectancy (EE), social influence (SI), facilitating condition (FC), hedonic motivation (HM), price value (PV), habit (H), trust (TR), privacy (PR), and behavioral intention (BI).
The results rejected H1, which hypothesized that performance expectancy would positively influence the users’ intentions to use the fruit and vegetable waste management platform. This result demonstrated an explicitly contradictory result with a standardized loading of 0.054 (p-value = 0.403).
H2 was rejected, which meant that effort expectancy had no effect on the users’ intentions to use the fruit and vegetable waste management platform. With a standardized loading of −0.012 (p-value = 0.857), this result showed a conflicting outcome.
H3 was supported, which recommended that social influence would positively influence the users’ intentions to use the fruit and vegetable waste management platform, with a standardized loading of 0.222 (p-value < 0.001). This study demonstrated that the inclination to use the platform was affected by social factors such as close friends, family, friends, and notable persons.
H4 was supported, which recommended that, with a standardized loading of 0.143 (p-value = 0.030 < 0.05), facilitating conditions would positively influence users’ intentions to use the fruit and vegetable waste management platform. This result was consistent with the hypothesis that users’ perceptions of convenience have a direct influence on their behavioral intentions to use technology. This indicates that consumers are motivated to use mobile shopping technologies when they have the necessary support from both the infrastructure and the technical side.
Moreover, the outcomes rejected H5, which stated that hedonic motivation does not influence the users’ intentions to use the fruit and vegetable waste management platform, with a standardized loading of 0.085 (p-value = 0.227).
The outcomes also rejected H6, which asserted that price value would positively influence the users’ intentions to use the fruit and vegetable waste management platform. With a standardization factor of 0.119 (p-value = 0.126), this result was contradictory.
H7 was supported, indicating that habit would positively influence the users’ intentions to use the fruit and vegetable waste management platform, with a standardized loading of 0.259 (p-value < 0.001). This study’s findings recommended that the routine activities of users had an effect on their intentions to use the platform.
H8 was supported, which hypothesized that trust would positively influence the users’ intentions to use the fruit and vegetable waste management platform, given a standardized loading of 0.158 (p-value = 0.046 < 0.05). The results of this survey showed that the users of the platform were quite conscious of the trust that is provided by the platform.
H9 was supported, which recommended that privacy would positively influence the users’ intentions to use the fruit and vegetable waste management platform, with a standardized loading of 0.365 (p-value < 0.001). This study demonstrated that the platform users had a high level of awareness regarding the safety of their personal data.
4.4. Multigroup Moderation Analysis
Measurement invariance (MI) is a method that utilizes indicators to determine a latent characteristic across all the respondents in Segments 1 and 2 [
121]. The MI technique has three parts: configural invariance, metric invariance, and scalar invariance. The distinction between full and partial MI is based on the findings from the following investigation. When only the conditions for configural invariance and metric invariance are met, partial invariance can be considered to exist. Despite this, full measurement invariance is established after both partial MI and scalar invariance are taken into account [
121].
According to
Table 8, all the CFI, IFI, and TLI values for configural invariance, metric invariance, and scalar invariance were high enough (>0.0900) to be considered an accepted fit [
126]. The findings demonstrated that the specified criteria were met by the remaining fit indices. The configuration invariance, metric invariance, and scalar invariance were acceptable. Thus, the full MI was constructed.
Table 9 displays the GOF measures and thresholds for the multigroup structural model. All the results were acceptable because they were above the minimum requirements that were specified. The comparative fit index (CFI; 0.923), incremental fit index (IFI; 0.924), Tucker–Lewis index (TLI; 0.912), and root mean square error of approximation (RMSEA; 0.049) passed the thresholds.
After the MGA was run, the hypothesis test results for Segments 1 (I) and Segment 2 (II) could be analyzed separately. According to
Table 10 and
Figure 3, H1 (I) and (II), H2 (I) and (II), H3 (II), H4 (I) and (II), H5 (I) and (II), H6 (II), H8 (II), and H9 (I) were not statistically significant, due to the
p-values of <0.05, <0.01, or <0.001. This insignificant relationship implies that, when separating the data into two segments, performance expectation, effort expectance, facilitating condition, and hedonic motivation did not influence the behavioral intentions to use the platform. This result occurred in both segments. However, H3 (I), H6 (I), H7 (I) and (II), H8 (I), and H9 (II) were accepted because their
p-values were less than the given level of significance. Additionally, the critical ratio difference revealed a significant path difference only for H7 (critical ratio = |−2.011| > |1.96|), indicating that the loadings of Segments 1 and 2 were statistically different. These results are discussed in more detail in the next section.
5. Discussion
Based on the statistical results from the SEM and MGA (Segment 1 vs. Segment 2) analyses, we can discuss the findings from the H1 to H9 tests as follows.
Hypothesis H1. Performance expectancy positively influences users’ behavioral intentions to use the circular agricultural waste management platform.
The findings of the study revealed that the respondents’ performance expectancy had no beneficial effect on their use of the platform (
p-value = 0.403 > 0.05). This was because “performance expectancy” refers to providing benefits and possibilities that are valuable for users in agricultural waste management, yet most of them did not express a high level of perception of this performance expectancy. Additionally, they may have felt that adopting a platform did not increase their productivity. Consequently, the majority of the respondents did not place a great deal of importance on performance expectancy. The study was not comparable to earlier research by Yeo et al. (2017) [
41], who found that performance expectancy was the extent to which users’ propensities to utilize online food-ordering systems were significantly influenced by the usability of the system.
Moreover, based on the results from MGA, the moderation effect of the demographic segments did not affect this relationship. The results revealed that neither Segment 1 (older and various incomes) nor Segment 2 (young and low incomes) moderated the relationship between performance expectancy and behavioral intentions, given the insignificant
p-values of Segment 1 (
p-value = 0.288 > 0.05) and Segment 2 (
p-value = 0.852 > 0.05). In this respect, performance expectancy did not influence the intentions to use an agricultural waste management platform for the users of all ages and income statuses. This was contradictory to the previous findings of Yusof et al. (2019) [
98], Bilgihan (2016) [
106], and Naruetharadhol et al. (2022) [
109].
Hypothesis H2. Effort expectancy positively influences users’ behavioral intentions to use a circular agricultural waste management platform.
In line with findings from earlier research, this study’s findings (
p-value = 0.857 > 0.05) revealed that the level of expected effort did not have any impact on the platform utilization [
9,
44], because most of the respondents were unfamiliar with the technology. As a result, they may have found it difficult to utilize this platform, the interactions between the users and this platform may have been confusing, and the functioning was still problematic. Therefore, the respondents were not concerned about the effort expectancy of using the platform.
The MGA result revealed that the moderation effect of the demographic segments did not affect this relationship either. The findings indicated that the association between effort expectancy and behavioral intentions was not moderated by either Segment 1, which comprised individuals with diverse incomes and advanced ages, or Segment 2, which consisted of young individuals with low incomes. This was evidenced by the non-significant
p-values of Segment 1 (
p-value = 0.464 > 0.05) and Segment 2 (
p-value = 0.474 > 0.05). It can be observed that the intention to utilize an agricultural waste management platform was not affected by the level of effort expectancy among the users in both segments. This assertion is in contrast with the earlier technology adoption research conducted by Yusof et al. (2019) [
98] and Naruetharadhol et al. (2022) [
109].
Hypothesis H3. Social influences positively impact behavioral intentions to use the agricultural waste management platform.
This gives credibility to the positive benefits that social influence has on behavioral intention (
p-value < 0.001), which have been found in earlier research [
21]. The result of this research implied that the information obtained from others around someone adds to their social influence. This influenced the inclination to use the platform more often, because the respondents may have been interested in a platform for disposing of agricultural waste management. This would also be a way for merchants and farmers to make more money and is expected to have a big impact on society.
Based on the MGA result, social influence (H3) appeared to affect the behavioral intentions in Segment 1 (loading = 0.251,
p-value = 0.042 < 0.05), but was irrelevant for Segment 2 (loading = 0.159,
p-value = 0.052). This result implies that the users in Segment 1 (older and various incomes) needed to have influences from their peers or third parties prior to actual use, while the users in Segment 2 (young and low incomes) did not need this. Because the older users in Segment 1 were not technology-oriented [
98], social influence would play a critical role in attracting their intention.
Hypothesis H4. Facilitating conditions positively influence users’ behavioral intentions to use an agricultural waste management platform.
This conclusion from the structural model revealed that previous research on the acceptability of the platform’s facilitating conditions had a major influence on the users’ intentions, given the
p-value of 0.03 (<0.05 level of significance). This finding supports Lu et al. (2005) [
52] and Naruetharadhol et al. (2023) [
9]. However, when strictly analyzing this relationship using a
p-value of 0.01, this hypothesis can be rejected and would lead to unsupported results for MGA.
However, the result of MGA revealed that the facilitating condition did not affect the userss behavioral intentions in both Segments 1 and 2, given the
p-values of 0.139 (>0.05) and 0.134 (>0.05), respectively. Most of the respondents believed the platform having customer services to help customers and provide information was not necessary. They also expected that the platform would not be compatible with the other technologies that they used. Hence, the MGA results contradicted Lu et al. (2005) [
52] and Naruetharadhol et al. (2023) [
9].
Hypothesis H5. Hedonic motivation positively influences users’ behavioral intentions to use an agricultural waste management platform.
The research results showed that the hedonic motivation of the respondents did not impact their intentions to use the platform (
p-value = 0.227 > 0.05), which differed from previous studies by Venkatesh et al. (2012) [
36] and Naruetharadhol et al. (2022) [
109], wherein hedonic motivation was shown to be a crucial component of behavioral intentions to adopt mobile internet in a consumer setting.
The MGA result also revealed that demographic segments did not moderate this structural relationship. This was because most of the respondents assumed that the platform was not focused on entertainment or enjoyment enough to affect the users’ intentions. According to the findings, the demographic segments did not have an impact on the relationship between hedonic motivation and behavioral intentions. Specifically, Segment 1, which included individuals with varying income levels and an older age (
p-value = 0.681 > 0.05), and Segment 2, which was made up of younger individuals with lower incomes (
p-value = 0.231 > 0.05), did not affect this association. These MGA results are unsupported by Venkatesh et al. (2012) [
36] and Naruetharadhol et al. (2022) [
109].
Hypothesis H6. Price value positively influences users’ behavioral intentions to use an agricultural waste management platform.
The findings of the study revealed that respondents’ price values had no effect on their use of the platform (
p-value = 0.126 > 0.05). Users may have been uninterested in the platform because they perceived it to be a relatively new platform (farmers and sellers), making it difficult to generate money. Some responders may also have been concerned about the expenses associated with utilizing this platform. For this reason, this hypothesis had no significant effect compared to previous research [
40].
When using the MGA approach, nevertheless, price value (H6) highly influenced the behavioral intentions in Segment 1 (loading = 0.401,
p-value = 0.005 < 0.001), but not in Segment 2 (loading = 0.022,
p-value = 0.827).
Table 10 also indicates a critical ratio difference of |−2.011|, which is more than the threshold of |1.96|, confirming that the loading of Segment 1 was statistically greater than that of Segment 2. This reveals that the users in Segment 1 tended to pay attention to price value because they expected the platform to be reasonably priced to attract their behavioral intention [
59]. On the other hand, the users in Segment 2 were not more likely to be concerned about price value.
Hypothesis H7. Habit positively influences users’ behavioral intentions to use an agricultural waste management platform.
The findings of the SEM were consistent with previous studies by Venkatesh et al. (2012) [
36], which discovered that habit affected behavioral intentions (
p-value < 0.001). This research identified that habit directly affected the behavioral intentions toward using the platform. Due to the platform information given to these potential users before each survey, most respondents are expected to use this platform and the platform is expected become a habituation for users to sell agricultural waste.
Furthermore, when considering the MGA result, habit affected the behavioral intentions in both Segment 1 (loading = 0.299, p-value = 0.020 < 0.05) and Segment 2 (loading = 0.182, p-value = 0.029 < 0.05). The impact was slightly higher in Segment 1 than in Segment 2. This shows that the users in both segments expected to use the platform often or as a part of their regular activities. This platform would be their first choice when it comes to working.
Hypothesis H8. Trust positively influences users’ behavioral intentions to use an agricultural waste management platform.
The SEM result suggested that trust robustly enhanced the behavioral intentions to use a platform for agricultural waste management (
p-value = 0.046 < 0.05). According to the survey results, both the farmers and sellers felt that the usage of this platform was dependable, that it maintained its promises to its users, and that the material on this platform was reputable. This is consistent with the earlier literature, which has demonstrated that trust has a noticeable effect on behavioral intentions [
37].
The MGA result suggested that trust influenced the behavioral intentions in Segment 1 (loading = 0.304, p-value = 0.032 < 0.05), but not in Segment 2 (loading = 0.059, p-value = 0.575). The test result demonstrated that trust may have better influenced the users in Segment 1 (older users) than in Segment 2 (younger users).
Hypothesis H9. Privacy positively influences users’ behavioral intentions to use an agricultural waste management platform.
The results indicated that privacy had a great impact on the users’ behavioral intentions to use the platform (
p-value < 0.001), comparable to prior research from Yuduang et al. (2022) [
79]. According to the survey findings, the respondents believed that the privacy of this platform’s users was protected and that the platform would keep the information of the participants safe. These findings are encouraging, because respondents’ perceptions of privacy are a significant factor in their decision to use a platform.
For the MGA result, privacy also did not impact the behavioral intentions in Segment 1 (loading = 0.256, p-value = 0.063), but strongly affected them in Segment 2 (loading = 0.394, p-value = ***). Because the respondents in Segment 1 were mostly from Generation X and baby boomers, they may not place as much emphasis on data privacy in their use of online platforms as the respondents in Segment 2 do, which was made up of people in the age range with unlimited internet access, combined with the development of technologies and social media.
8. Conclusions
The purpose of this study was to identify the different demographic segments of users, as well as the significant factors that influence the behavioral intentions to utilize a circular-economy-based digital platform for fruit and vegetable waste management, based on the Unified Theory of the Adoption and Utilization of Technology 2 (UTAUT2). The study found two user segments: (1) older and various-income users, and (2) young and low-income users. Our data suggest that social influence, price value, habit, trust, and privacy are essential elements impacting these user behavioral intentions when using a fruit and vegetable waste management platform. Contrary to the UTAUT2, there was no substantial relationship between performance expectancy, effort expectancy, facilitating condition, hedonic incentive, and behavioral intentions towards adopting the technology. The results show that the attitudes and behaviors of surrounding society influence consumers’ adoption of waste management platforms, especially for older and various-income users. When motivated by a group of people who influence beliefs or feelings, such as a colleague or a person who has the same or a similar occupation, the more probable it is that the Segment 1 users would be driven, convinced, and followed by them. Additionally, the results revealed that price value tended to influence the users in Segment 1, but did not affect those in Segment 2. Moreover, the data revealed that consumers wanted to engage in online behavior owing to habit in both segments, as familiarity influenced the propensity to use the platform. Its intended use is equally vital to directing its usage behavior. Therefore, the platform developers should promote the benefits of its use and encourage continuing use by, for example, customer service, a user-friendly and straightforward platform, or more services. Furthermore, the results demonstrated that users have online behavioral intentions due to trust. Explicitly, trust increased interest in using the platform for Segment 1 users. This might establish credibility and influence behavioral intentions. Hence, the platform developers should provide clear information and contact options to increase the platform’s credibility. Additionally, the results demonstrated that users have online behavioral intentions due to privacy, especially the younger users in Segment 2; thus, the platform developers should create the most secure platform possible to safely store user information without publishing it. Therefore, this study suggests that the platform developers should promote users’ behavioral intentions differently through specific target groups for fruit and vegetable sellers and farmers from various age and income groups.